Studies were eligible if they satisfied the following criteria: a) Participants: adults (≥18 years) with coronary artery disease (i.e., history of coronary artery disease with angina pectoris or myocardial infarction diagnosed based on the American Heart Association standard criteria,14 angiographically documented cases, percutaneous coronary intervention, or coronary artery bypass grafting); b) Intervention/comparison: studies comparing HIIT (i.e., performed at 70–85 % peak VO2, 75–90 % maximal heart rate, 70–85 % heart rate reserve, or 14–16 rating on the Borg scale) and MICT (i.e., performed at 40–69 % peak VO2, 55–74 % maximal heart rate, 40–69 % heart rate reserve, or 12–13 rating on the Borg scale) that were completed on a bike or on a treadmill and performed at least twice a week for at least four weeks; c) Outcomes: VO2peak (measured in mL/kg/min by cardiopulmonary exercise test) and HRQoL (measured by any standardized and validated scales or questionnaires); d) Study design: an RCT. Studies that enrolled patients with other cardiac or respiratory diseases were excluded.

We searched for publications in MEDLINE/PubMed, the Cochrane Library, EMBASE, and the PEDro databases up to December 2022 without language restrictions. We used a standardized search strategy and, whenever possible, the same search terms (e.g., MeSH term for PubMed and the Cochrane Library and EMTREE for EMBASE). Our search strategy included three groups of keywords and their synonyms: study design, participants, and interventions.

The strategy developed by Higgins et al.15 was used to identify the RCTs in MEDLINE/PubMed. To identify the RCTs in other databases, we adopted a search strategy using similar terms. The references of the articles included in this meta-analysis were checked to identify other potentially eligible studies. The search strategies for MEDLINE/PubMed, EMBASE, and the Cochrane Library are presented in Supplementary Material - S1. For ongoing studies, confirmation of any data, or additional information, the authors were contacted through e-mail.

Titles and abstracts were independently checked by two reviewers. If at least one of the reviewers considered a reference eligible, the full text was obtained for complete assessment. Two reviewers independently assessed the full texts of the selected articles to verify whether they met the eligibility criteria. Two authors independently extracted data from the published reports using standard data extraction forms adapted from Higgins et al.15 The study population, intervention performed, follow-up period, rates of missing data, outcome measures, and results were reviewed.

The analysis of the methodological quality of the included studies was carried out using the PEDro scale.16 This scale consists of 11 items (1: eligibility criteria and source of participants; 2: random allocation; 3: concealed allocation; 4: baseline comparability; 5: blinded participants; 6: blinded therapists; 7: blind assessors; 8: adequate follow-up; 9: intention-to-treat analysis; 10: between-group comparisons; and 11: point estimates and variability). However, one item (eligibility criterion) is related to external validity and is generally not used to calculate the score, resulting in a score range of 0–10.16 Initially, the classification already available in the database was used. If the study was not yet classified, two authors carried out the analysis and classified the studies.

Effect estimates were obtained by comparing the mean change from baseline to the endpoint for each group. Thus, change scores [i.e., post- minus pre-intervention scores] were extracted from each study and expressed as mean ± standard deviation (SD). When the outcome change score was not described in the study, we made the calculation. To calculate the standard deviation of change we used the statistics (confidence intervals, SEs, t statistics, P values, F statistics) presented in the study according to recommendations made in the Cochrane Handbook chapters 6 and 10 using revman.15 In cases where studies presented data as mean ± standard error of the mean (SEM), SEM was converted to SD.17

The conversion of nonparametric data (i.e., median and interquartile range) to means and SDs was performed using recently established methods.18 In cases where the SD of change was unavailable, but confidence interval (CI) was available, we converted the CI to SD using the method described by Higgins et al.15 If the required outcome data were not found in the full text but were presented graphically, an attempt was made to digitize the graph.19

We compared HIIT versus MICT and performed a sub-analysis of studies that compared isocaloric exercise training protocols. Isocaloric protocols aim to adjust the energy expenditure within aerobic exercise sessions performed with different intensities.11 We considered isocaloric protocols when reported in the studies.

An α value of 0.05 was considered significant. Calculations were performed using a random-effects model. Statistical heterogeneity of the treatment effect among studies was assessed using Cochran's Q test and the I2 inconsistency test statistic. In these tests, values of 0 % to 40 % might not be important; 30 % to 60 % may represent moderate heterogeneity; 50 % to 90 % may represent substantial heterogeneity; and 75 % to 100 % may indicate considerable heterogeneity.15 All analyses were conducted using Review Manager version 5.4 (Cochrane Collaboration).20

The sensitivity analysis was performed by re-running the meta-analysis, excluding studies with a high risk of bias according to PEDro score (≤4 points), and then employing the leave-one-out method by removing one study at a time and repeating the analysis.

The certainty of evidence was assessed using the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) approach. Quantitative data from Cochrane's review manager were imported into GRADEpro GDT 2015 to create a “Summary of findings table.” The assessment involved five items: risk of bias, imprecision, inconsistency, indirectness, and publication bias.15,21 The certainty of evidence was downgraded by one level for risk of bias if more than a quarter of the studies included in the meta-analysis were considered to have a high risk of bias (i.e., defined as RCTs with a PEDro score ≤4 points). Imprecision was evaluated by visually analyzing the confidence intervals. If necessary, imprecision was evaluated using absolute risk analysis, number needed to treat, or optimal information size calculation. Results were considered imprecise if the total sample size was <300 and <400 for dichotomous and continuous outcomes, respectively. We downgraded for inconsistency if there was clinical or statistical heterogeneity (i.e., I2>40 %) or if there were no overlapping confidence intervals. For indirectness, we initially observed the pre-established PICO criteria (Population, Intervention, Comparator, and outcome). We compared the characteristics of the study populations to those of the population of interest outlined in our PICO criteria. Discrepancies such as age, disease severity, and comorbidities were recorded. We evaluated the interventions and comparators in each study for concordance with those defined in our review protocol. We examined whether the studies reported outcomes that were directly relevant to our research question and whether these outcomes were measured and reported in a way that allowed comparison across studies. Publication bias was assessed by visual inspection of funnel plots (i.e., scatterplots of the effect size from individual studies against its standard error) for meta-analyses with at least 10 trials.15,21-23 Reasons for downgrading were noted and attached to the summary of findings table.

The initial search identified 892 titles and abstracts. Of these, 31 were considered potentially relevant and retrieved for full-text analysis. Finally, 27 studies24-50 involving 1454 patients met the eligibility criteria and were included in the systematic review. A flowchart depicting the selection process for each stage is shown in Fig. 1. PEdro scale scores of the included studies are presented in Supplementary Material - S2.

Fig. 1.

PRISMA flow chart.

(0.34MB).

The mean age of the participants ranged from 50 to 70 years. Twenty-three studies included patients of both sexes and four studies included only men. The sample size, population characteristics, outcomes, and characteristics of the exercise programs of the included studies are summarized in Supplementary Material - S3. HIIT programs were generally performed 2 to 4 times per week, with 4 × 4-minute high-intensity intervals, with intensities ranging from 85 to 95 % of peak heart rate and The Borg rating of perceived exertion scale (Borg scale) rating >15. The MICT programs were generally performed 2 to 4 times per week, with intensities ranging from 60 % to 70 % of peak heart rate and Borg scale rating of 11–14.

Twenty-four studies with 1259 patients assessed VO2peak as an outcome. The studies showed mean VO2peak at baseline and post-intervention of 24 mL/kg/min and 28 mL/kg/min, respectively. The meta-analysis showed that the HIIT group improved VO2peak by 2.11 mL/kg/min (95 % confidence interval [CI]: 1.14, 3.07, I2 = 78 %, N = 1259, low certainty of evidence) greater than the MICT group (Fig. 2).

Fig. 2.

Change in peak VO2 – HIIT versus MICT.

(0.51MB).

The meta-analyses of the 14 studies that did not use isocaloric exercise training showed that the HIIT group improved VO2peak by 2.94 mL/kg/min (95 % CI: 1.44, 4.44, I2 = 70 %, N = 806, low certainty evidence) compared to the MICT group (Fig. 2). When we removed one study at a time, the study by Benetti et al.47 had the most impact in reducing heterogeneity, which decreased from 70 to 0 %. The MD decreased from 2.94 to 2.26 mL/kg/min (95 % CI: 1.43, 3.08). When we also removed studies of low methodological quality (PEDro score of ≤4 points), heterogeneity was also reduced to 0 % and the MD decreased to 2.33 mL/kg/min (95 % CI: 1.38, 3.28, I2=0 %, N = 601).

Ten studies used isocaloric exercise training protocols. This subgroup analysis showed that HIIT may have little or no effect on VO2peak (MD= 0.72 mL/kg/min, 95 % CI: −0.03, 1.48, I2 = 44 %, N = 453, low certainty evidence) compared to the MICT group (Fig. 2). When we removed one study at a time, the study by Bouri et al.44 had the most impact in reducing heterogeneity, which decreased from 44 to 0 %. The MD decreased from 0.72 mL/kg/min to 0.15 mL/kg/min (95 % CI: −0.27, 0.57). When we also removed studies of low methodological quality (PEDro score ≤4 points), heterogeneity was also reduced to 0 % and the MD decreased to 0.21 mL/kg/min (95 % CI: −0.35, 0.76, I2=0 %, N = 365). Certainty of evidence is presented in Table 1.

Six studies involving 316 patients assessed the HRQoL using the MacNew questionnaire (global, physical, emotional, and social domains). No difference were found in all domains between participants in the HIIT group and those in the MICT group (Fig. 3). The certainty of the evidence was rated as very low (Table 1).

Fig. 3.

Change in quality of life (MacNew) – HIIT versus MICT.

(0.71MB).